Molecular interactions are the underlying basis of all processes that are executed in an organism, and their complete mapping would be a great aid in understanding and interpreting both normal and disease functioning. Transcriptional regulatory interactions are of particular interest as they are critical in the proper spatial and temporal regulation of genes. This proposal aims to develop several novel and complementary computational methods for predicting transcription factor interactions and specificities, and for uncovering their conservation and variation across organisms. Taken together, these methods will vastly expand our knowledge of eukaryotic regulatory networks and their underlying principles. We will devise a combined constrained optimization and statistical approach to predict the DNA-binding specificities of multidomain C2H2 zinc finger proteins; these proteins comprise the largest class of transcription factors in eukaryotic genomes. We will also establish a novel comparative sequence framework for determining binding specificity variation amongst homologous transcription factors, as network divergence underlies much of the observed phenotypic and functional diversity between and within organisms; this framework will be applied to explore the extent to which changes in transcription factors can affect regulatory network variation across organisms. Finally, we will develop a cross-genomic framework for predicting genomic binding sites for transcription factors with known specificities, along with analysis techniques for inferring interactions amongst these transcription factors across organisms. The DNA-binding specificities for an increasing number of transcription factors are being determined, and this large-scale data presents new opportunities to map transcription factor binding sites and to uncover transcription factor- transcription factor interactions across organisms; these interactions are an important component of regulatory networks and their variation plays a key role in network divergence. Successful completion of these aims will result in computational methods that will significantly increase the rate with which transcriptional networks are characterized and will reveal fundamental aspects of their functioning and evolution. All developed software will be made publicly available.